Impact of Glucose-Lowering Medications on Health-Related Quality of Life in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE)

Quality of life reflects general well-being and has been defined as “an individuals’ perception of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns” (1). Health-related quality of life (HRQoL) focuses more specifically on the aspects of quality of life that are affected by health (2). HRQoL is recognized as a multidimensional concept including domains of physical health and functioning, mental health, social functioning, and how they are influenced by a medical condition and its treatment (3). Diabetes has been associated not only with increased morbidity and mortality but also with reduced HRQoL (4,5).

Numerous studies have included examination of HRQoL among individuals living with diabetes (4,5). In a cross-sectional study by Tapp et al. (6), previously diagnosed diabetes was associated with a significantly greater risk of scoring in the lowest quartile of all but one dimension of the SF-36 scale. There was also evidence of reduced HRQoL on the general health, physical functioning, and role limitation physical scales of the SF-36 scale among those with newly diagnosed diabetes and the physical functioning and social functioning scales among those with impaired glucose tolerance (6). In a recent systematic review and meta-analysis investigators found that longer duration of diabetes, the presence of comorbidities (depression, hypertension), and diabetes-related complications were all associated with lower HRQoL (4).

However, once comorbidities and complications are taken into account, there are few studies examining the association of diabetes medications and their associated effects (hypoglycemia, weight changes, burden of treatment, etc.) with HRQoL, particularly in patients early in the disease process. To date, studies examining the effect of specific glucose-lowering medications have shown mixed results, in part owing to differences in study duration, drug combinations, and measures of HRQoL (7). There are no comparative effectiveness studies that have prospectively investigated changes in HRQoL in patients with type 2 diabetes (T2D) treated with metformin in combination with commonly used second-line agents. Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) was a comparative effectiveness study that enrolled individuals with T2D within 10 years of diagnosis on metformin monotherapy and then randomized participants to one of four glucose-lowering medications. The relatively early stage of disease, low baseline levels of comorbidities and complications, and randomization to treatment group allow for a rigorous assessment of changes in HRQoL between treatments. We also assessed whether those changes, if present, were mediated by changes in treatment-related effects.

GRADE was a multicenter, parallel-group, comparative effectiveness clinical trial. Additional details have previously been published (8–11). Eligible participants were randomly assigned to one of four glucose-lowering medications (1:1:1:1) in combination with metformin: glargine (insulin glargine U-100), glimepiride (sulfonylurea), liraglutide (glucagon-like peptide 1 [GLP-1] receptor agonist), and sitagliptin (dipeptidyl peptidase 4 [DPP-4] inhibitor) (8). The Consolidated Standards of Reporting Trials (CONSORT) diagram is included in Supplementary Fig. 1. Eligibility requirements for GRADE at screening and randomization have previously been reported (8–10). Eligible participants had T2D with <10 years’ duration, were diagnosed at age ≥30 years for non-American Indian/Alaska Native or age ≥20 years for American Indian/Alaska Native, were taking metformin monotherapy (at least 1,000 mg/day), had HbA_1c 6.8%–8.5% (51–69 mmol/mol) at randomization, and were willing to take a second diabetes medication, including daily injections of insulin or liraglutide. Over a 4-year period, 5,047 participants were enrolled and then followed for an average of 5 years. The metabolic outcome in GRADE was HbA_1c ≥7% (53 mmol/mol) at a quarterly visit that was subsequently confirmed; the secondary metabolic outcome was a value >7.5% (58 mmol/mol), confirmed. On reaching the secondary GRADE outcome, those assigned to glargine were to initiate “rescue” insulin with the addition of aspart, whereas those in the other groups were to initiate glargine. All participants who completed any follow-up HRQoL assessments were included in these analyses and descriptions in this article.

Scores on the self-administered version of the Quality of Well-being Scale (QWB-SA), the SF-36 mental component summary (MCS) scale, and the SF-36 physical component summary (PCS) scale served as coprimary outcomes. Scores on the eight SF-36 subscales were analyzed as secondary outcomes, including physical (general health, physical functioning, role limitations due to physical health problems, and bodily pain) and mental (mental health, social functioning, role limitations due to personal or emotional difficulties, and vitality) components. The QWB-SA is a preference-based measurement tool with a combined assessment of symptoms (acute and chronic) and functioning (mobility, physical activity, and social activity) to generate a health utility score as a measure of HRQoL (12). It has been used to examine HRQoL in populations and subgroups with a number of different medical conditions, including T2D (13).

Linear regression models were used to estimate the mean changes from baseline separately for each outcome (PCS and MCS, each of the eight SF-36 subscales, and QWB-SA) and separately at years 1, 2, and 3. Each was also adjusted for the baseline value of the HRQoL measure (intention-to-treat [ITT] models) and then adjusted for insulin initiation at that year and separately for weight change from baseline to that year. Similar models were used to estimate the changes from baseline to years 1, 2, and 3 in the secondary outcome analyses of the subscales of the SF-36: bodily pain, general health, mental health, physical functioning, role limitations (emotional), role limitations (physical), social functioning, and vitality. For display of mean values longitudinally, identical models were applied to the actual measure at each year to provide estimates of the longitudinal mean values (rather than changes from baseline). Since both analyses (change or follow-up value) are adjusted for the baseline value of HRQoL, the significance level of each is identical. The P values for the overall test of any treatment differences were adjusted with the Holm method for three tests for the primary outcomes (QWB-SA, MCS, PCS).

Prespecified subgroup analyses were performed for assessment of whether the pattern of outcome means was homogeneous among strata for the following: age-groups (<45, 45–59, ≥60 years), sex (female, male), race (Black, White, other), ethnicity (non-Hispanic, Hispanic), BMI tertiles (18.2–30.7, 30.8–36.2, 36.3–74.3 kg/m²), standard BMI categories (18 to <25, 25 to <30, ≥30 kg/m²), baseline HbA_1c (6.8–7.2%, 7.3–7.7%, 7.8–8.5% or 50.8–55.2, 56.3–60.7, 61.8–69.4 mmol/mol), diabetes duration tertiles (0–2.3, 2.4–5.2, 5.3–10.9 years), history of myocardial infarction or stroke (yes, no), history of kidney disease (yes, no), and history of retinopathy (yes, no).

For a given factor (e.g., sex), the above longitudinal models were augmented by the addition of the subgroup factor and its interaction with treatment group as covariates in the models. The global test of homogeneity for that factor is a 3 × (S − 1) df χ² contrast test of homogeneity, where S is the number of strata and df is degree of freedom (14). For a given outcome, the P values for the tests of homogeneity for the 11 subgroup factors are adjusted to control the false discovery rate (11), a method that is less conservative (i.e., more powerful) than a Bonferroni-type step-down adjustment. If the global test is significant for a given subgroup factor, then six additional (S − 1) df tests of homogeneity for the six pairwise treatment comparisons are conducted and Holm corrected for six tests. If heterogeneity is detected for a given pairwise comparison (i.e., both the global and pairwise test of homogeneity are rejected), then the pairwise treatment differences of the outcome are estimated within each stratum and the P values for the test of the differences equal to zero are false discovery rate adjusted for S comparisons.

Potential mediators were selected a priori based on previous evidence of their association with HRQoL and include weight change (15) and initiation of insulin treatment (5) as well as changes in HbA_1c (16), nonsevere hypoglycemia (defined as two or more episodes of hypoglycemia requiring food or medication over the last 30 days), gastrointestinal symptoms (defined as having weekly or daily occurrences of at least one of nausea, vomiting, bloating, or diarrhea over the last 30 days), and the number of reported incident serious adverse events, which included episodes of severe hypoglycemia.

Mediation is calculated as the percentage of treatment effects explained by mediators, and it is estimated as the percent change in the treatment effect coefficient from a model without the mediator included as a covariate versus a model that includes the mediator as a covariate (17). The estimate of the percentage of treatment effect explained by the mediator and its 95% CI are presented.

Data completion rates for the outcomes were high at all visits (90%–95%) and did not differ by treatment group. No special missing data methods (e.g., imputation, sensitivity analyses) were used because the results were unlikely to be biased given the completeness of the data.

R, version 4.2.1, was used for all analyses (18).

This article is based on follow-up data and outcome assessments from the 5,047 participants enrolled into GRADE, the database for which will be available in the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Central Repository in 2024.

Baseline characteristics of GRADE participants (n = 4,885) are presented in Supplementary Table 1. On average, participants were 57 years old, and 34%–38% were female, 17%–21% were Black or African American, and 17%–19% were Hispanic or Latino. Average duration of diabetes was ∼4 years and BMI 34.3 kg/m², with mean HbA_1c 7.5% (58.5 mmol/mol). Apart from treatment group differences in fasting insulin and insulinogenic index, all other baseline characteristics were equally distributed in the four treatment groups. There were no significant differences across groups in baseline measures of HRQoL.

Mean weight (kilograms) for each treatment group from quarterly visits through year 3 is shown in Supplementary Fig. 2. Participants in the glargine group had little change in weight over time, the glimepiride group had an early rise through 6 months and then a gradual decline to 3 years with little overall difference between baseline and 36 months, the sitagliptin group decreased steadily throughout follow-up, and the liraglutide group decreased rapidly to a nadir at 9 months and then remained stable through 3 years. The number of participants initiating rescue insulin before the 1, 2, and 3 year study visits in each of the treatment groups is shown in Supplementary Table 3. Also shown in Supplementary Table 3 is the average time from baseline until the initiation of rescue insulin. The lowest rates of rescue insulin initiation occurred in the glargine group and the highest rates occurred in the sitagliptin group, with rates for the glimepiride and liraglutide groups between these extremes.

The mean changes in primary HRQoL outcomes at years 1, 2, and 3 are shown in Table 1 for the ITT analysis and for analyses adjusted for weight change and initiation of rescue insulin. Overall, there were modest but significant within-group improvements for HRQoL outcomes for all four treatment groups at various time points. Notably, there were no significant reductions in HRQoL outcomes in any treatment group. Among the primary outcomes in the current study, there was a significant ITT difference among treatment groups in mean SF-36 PCS score (P < 0.05, data not shown) at year 1. After adjustment for six pairwise comparisons, the liraglutide group had significantly greater change in physical function (1.52-point increase) compared with both the glargine and glimepiride groups (0.63 increase for each; adjusted P < 0.05 for both). The differences between the liraglutide group compared with glargine and glimepiride groups at 1 year remained significant after adjustment for insulin initiation (liraglutide increase of 0.70 vs. decreases of 0.21 and 0.20 in the glargine and glimepiride groups; P < 0.05 for both comparisons) but were no longer significant after adjustment for weight change at 1 year (liraglutide increase of 1.20 vs. increases of 0.83 and 0.85 for glargine and glimepiride; P > 0.05 for both comparisons). Pairwise treatment comparisons for PCS change scores at year 1 similarly demonstrated significant differences for liraglutide compared with glimepiride and glargine; there were no other significant differences between medications (data not shown). Figure 1 depicts the mean change in each of the three HRQoL measures over time by treatment group. There were no significant ITT differences among treatments for either QWB-SA or MCS at 1 year. There were no significant ITT differences among treatments for any primary outcome at years 2 and 3.

Mean changes in secondary outcomes (SF-36 eight subscales) at years 1, 2, and 3 are shown in Supplementary Table 2. There were no significant treatment group differences in the secondary outcomes at any of the three time points.

The results of the mediation analysis for the change in PCS score at 1 year are shown in Supplementary Table 4. Only weight change was a significant mediator of the liraglutide versus glargine, and liraglutide versus glimepiride, differences described in Table 1. Weight change explained 52.8% (95% CI 14.7–90.9) of the liraglutide versus glargine treatment effect and 56.7% (95% CI 15.9–97.5) of the liraglutide versus glimepiride treatment effect.

Subgroup analyses (Supplementary Table 5) showed that the treatment differences in HRQoL were driven by differences in the subgroup of participants in the highest BMI tertile at baseline (36.3–74.3 kg/m²). Figure 2 shows the mean change of PCS over time in the BMI subgroups where all participants in each BMI subgroup were assumed to start from a common mean PCS score at baseline. Comparison across the three panels shows the differences in average baseline PCS scores between the subgroups, with tertile 1 representing the highest PCS score and tertile 3 the lowest. Note that the difference from baseline to year 1 represents the mean change in PCS score at 1 year presented in Supplementary Table 5. The panels in Fig. 2 representing the three subgroups show that the treatment groups are acting very similarly for the first and second BMI tertile subgroups but there is a larger increase in PCS score in the liraglutide group than in the other three treatment groups within the highest BMI tertile subgroup. This difference did not persist at years 2 and 3. Apart from BMI, additional subgroup analyses by participant demographics (age, sex, race/ethnicity), diabetes duration, or presence of comorbidities (history of myocardial infarction, kidney disease, retinopathy) demonstrated no heterogeneity (data not shown).

In this large comparative effectiveness trial of four commonly used glucose-lowering medications from separate classes (insulins, sulfonylureas, GLP-1 receptor agonists, and DPP-4 inhibitors) added to metformin, there were no significant reductions in HRQoL indicators in any of the treatment groups, composed of participants with relatively brief duration of diabetes and with few complications. There were treatment group differences for the SF-36 PCS at year 1 but not at years 2 or 3. Liraglutide was associated with a modest increase in PCS (1.52) compared with glargine and glimepiride (0.63). Examination of subscales comprising the PCS suggests that this difference is primarily driven by reduced limitations in physical functioning, such as limitations in vigorous activities, carrying groceries, and climbing stairs. Roughly half of the improvements in PCS for the liraglutide group compared with the glimepiride and glargine group were attributable to weight loss. Improvements in PCS differed by baseline weight status. Those in the upper tertile of BMI (BMI >36.3 kg/m²) demonstrated the greatest improvements. Treatment group differences remained significant after adjustment for initiation of rescue insulin.

This study adds to the existing literature examining the association between glucose-lowering medications and HRQoL and uniquely allows for direct comparisons across different treatments. In 2016, Reaney et al. (7) conducted a systematic review of studies examining the impact of several classes of diabetes medications (GLP-1 receptor agonists, novel insulins, sodium–glucose cotransporter 2 inhibitors, DPP-4 inhibitors) on patient-reported outcomes including HRQoL. Several studies demonstrated limited to no significant change from baseline in HRQoL as measured with the EuroQol-5D (EQ-5D) and SF-36, including studies examining exenatide, liraglutide, dulaglutide, dapagliflozin, and insulin glargine. Some studies documented improvements in HRQoL in specific scenarios, such as with the addition of exenatide, dulaglutide, and sitagliptin to metformin (as measured with the EQ-5D) or for degludec in comparison with insulin glargine (as measured with the SF-36 PCS). Investigators of other studies observed worsening HRQoL, such as for dulaglutide compared with insulin glargine, both in combination with lispro (with or without metformin). Differences in drug combinations, varied length of follow-up, the use of numerous measurement instruments (n = 20), and the lack of comparative effectiveness studies prohibited authors from drawing overarching conclusions for class effects on HRQoL. In the current comparative effectiveness study, liraglutide was associated with a 1.5-point improvement in SF-36 PCS score at 1 year, 0.83 points more than the improvement with glimepiride or glargine. Previous work has demonstrated that for patients with diabetes, a 1-point decrement on selected SF-36 scales is associated with an excess risk of up to 9% for mortality and 12% for inability to work (19), suggesting a 1.5-point improvement is clinically meaningful (20). The current finding is also consistent with those of two recent large trials with examination of the association between GLP-1 receptor agonists and HRQoL. Liraglutide Effect and Action in Diabetes: Evaluation of Cardiovascular Outcome Results (LEADER) (21) was a randomized placebo-controlled trial examining cardiovascular (CV) outcomes in patients with T2D at high risk for cardiovascular disease (CVD). At 36 months, they observed less deterioration in HRQoL as measured with the EQ-5D in the liraglutide group compared with the placebo group. Differences were primarily driven by improvements in mobility and self-care (both domains within EQ-5D). In LEADER, weight loss of 5% demonstrated no impact on HRQoL, while initiation of insulin was associated with a modest reduction in HRQoL. The Trial to Evaluate Cardiovascular and Other Long-term Outcomes With Semaglutide in Subjects With Type 2 Diabetes (SUSTAIN-6) (22) was a large (n = 3,297) randomized, placebo-controlled trial testing the impact of semaglutide on CV outcomes in patients with T2D and age ≥50 years with established CVD or age ≥60 years with at least one CV risk factor. At 104 weeks, improved scores were found for the semaglutide group compared with placebo for the SF-36 PCS (1.0 vs. 0.4) and the SF-36 MCS (0.5 vs. −0.2), respectively. The treatment effects for semaglutide for both the PCS and the MCS were modestly reduced after controlling for change in HbA_1c; the treatment effect was further reduced for the PCS (but not for MCS) after adjustment for body weight. SUSTAIN-6 and LEADER both included patients with T2D with greater baseline morbidity than in GRADE, as the participants were at high risk for or had CVD. Both studies included comparison of treatment with placebo. GRADE extends the current evidence to support the association between use of GLP-1 receptor agonists and improved HRQoL even in patients within 10 years of diabetes diagnosis, with few complications and at lower risk of CVD. Most importantly, GRADE included active comparator treatments. Of note, the differences in HRQoL between the treatment groups did not persist beyond the first year.

In the current study, changes in PCS were largely driven by changes in weight, and more specifically, by changes in weight for those in the uppermost BMI tertile. Obesity is known to be associated with reduced HRQoL, especially in physical component scores (23). Previous weight loss studies have often demonstrated improvements in HRQoL, especially the studies testing intensive lifestyle and bariatric surgery interventions (24,25). However, the finding has been less consistent in studies of nonsurgical interventions, with variability due to multiple factors including intervention differences (medication vs. lifestyle), varying degrees of weight loss, differences in length of follow-up (ranging from 6 weeks to ≥24 months), and differing measures of HRQoL (generic vs. obesity specific). Like SUSTAIN-6, which showed reduced treatment effects on PCS after adjustment for weight, treatment effects on PCS at 1 year in GRADE were also rendered nonsignificant after inclusion of change in weight in the model. Further subgroup analyses demonstrated that treatment effects were largely driven by change in weight for those individuals in the upper tertile of BMI (>36.3 kg/m²). These potential HRQoL benefits further support the current American Diabetes Association Standards of Care in Diabetes (26) recommendation of selection of a GLP-1 receptor agonist when weight loss is a priority.

Unlike SUSTAIN-6, there were no treatment differences in GRADE for MCS or QWB-SA. This lack of benefit may be partially due to study design. SUSTAIN-6 was a randomized, placebo-controlled, efficacy study, whereas GRADE was a comparative effectiveness study with four active glycemia-lowering treatments. While there were no treatment differences for MCS or QWB-SA in GRADE, scores for both measures improved slightly within each of the four treatment groups. Participants in GRADE attended regular quarterly visits and could reach out to study staff for help as needed over a 5-year study period. It is possible that the supportive care provided in the clinical trial resulted in improvements across the board and may have masked small treatment differences. The lack of significant treatment differences for MCS and QWB-SA in GRADE suggests that treatment decisions regarding the addition of a second medication after metformin for patients like those enrolled in GRADE (i.e., within 10 years of diagnosis and at low risk for CVD) can be made without concern for any deleterious impact on HRQoL.

Strengths of this study include a relatively large, well-characterized, and diverse study cohort, with substantial representation of Black and Hispanic or Latino participants from across a range of socioeconomic strata recruited from all regions of the U.S. Rigorous study design with randomization to treatment groups and repeated measures of HRQoL are additional strengths. The study also has several limitations. The GRADE sample was composed of volunteers, restricted by the inclusion and exclusion criteria, and results cannot be extrapolated to the wider population with T2D. People likely enrolled in GRADE to improve diabetes management, and most did this successfully; this may have contributed to a general sense of well-being and mitigated small negative effects of medications on HRQoL measures. There is also the potential for misclassification of HRQoL outcomes, based on timing of HRQoL measures, and this potential bias could impact the results in either direction. Underestimation of HRQoL improvement could occur if the HRQoL measure occurred very shortly after an unfavorable event (e.g., hypoglycemia), while overestimation could occur if the HRQoL measure occurred soon after a favorable event (e.g., weight loss) occurred. Other measures of behavioral and psychosocial domains that may have impacts on HRQoL such as emotional state, mood, locus of control, and social support were not included as covariates in the analyses, limiting our ability to comment on the role of these factors. All of the medications in GRADE were provided free of charge, so the potential effects of the differential costs of the medications on HRQoL cannot be assessed. Further, GRADE did not include some of the newer glucose-lowering medications (such as weekly GLP-1 receptor agonists or sodium–glucose cotransporter 2 inhibitors), which may have differentially impacted HRQoL. Finally, this study assessed short term-changes in HRQoL. The enduring impact on quality of life assessment between therapies would require a longer study in order to include changes in diabetes-related complications over time, which are known to impact HRQoL.

In this study of 4,885 participants with T2D within 10 years of diagnosis at the time of randomization, none of the four treatments added to metformin had a deleterious effect on HRQoL. There were no treatment differences by other measures of HRQoL (SF-36 MCS and QWB-SA) at any time point. Those treated with liraglutide and metformin showed improvement in physical functioning as measured with the SF-36 PCS, which appeared to be related to weight loss and baseline BMI. Improvements were no longer significant at year 2 or 3. These findings add to a growing literature demonstrating the potential benefits of GLP-1 receptor agonists, albeit transient, on HRQoL.

Funding. GRADE was supported by a grant from the NIDDK of the National Institutes of Health (NIH) under award no. U01DK098246. The planning of GRADE was supported by a U34 planning grant from the NIDDK (U34-DK-088043). The American Diabetes Association supported the initial planning meeting for the U34 proposal. The National Heart, Lung, and Blood Institute and the Centers for Disease Control and Prevention also provided funding support. The Department of Veterans Affairs provided resources and facilities. Additional support was provided by NIH grants P30 DK017047, P30 DK020541-44, P30 DK020572, P30 DK072476, P30 DK079626, P30 DK092926, P30 DK111022, U54 GM104940, UL1 TR000439, UL1 TR000445, UL1 TR001108, UL1 TR001409, UL1 TR001449, UL1 TR002243, UL1 TR002345, UL1 TR002378, UL1 TR002489, UL1 TR002529, UL1 TR002535, UL1 TR002537, and UL1 TR002548. Educational materials were provided by the National Diabetes Education Program. Material support in the form of donated medications and supplies was provided by Becton, Dickinson and Company, Bristol-Myers Squibb, Merck, Novo Nordisk, Roche Diagnostics, and Sanofi.

The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. A.L.C. reports participation on an advisory board for Bright Bodies outside the submitted work. W.H.H. reports paid participation on a data and safety monitoring board or advisory board for Merck Sharp & Dohme (ertugliflozin) outside the submitted work. J.S.G. reports consulting fees from Virta Health; participation on a data and safety monitoring board or advisory board for the Mobile Phone Support for Adults and Support Persons to Live Well with Diabetes study, NIH grant 5R01DK119282-04, and receipt of materials or equipment (Abbott FreeStyle Libre) outside the submitted work. J.S.G. reports payment or honoraria for virtual presentation from the Worldwide Initiative for Diabetes Education, support for meeting attendance and/or travel from the American Diabetes Association (ADA) Professional Conference Planning Committee, and a leadership role as planning committee chair for ADA behavioral medicine outside the submitted work. None of these payments of support were made in relation to GRADE or to the development of this article. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. All authors affirmed that authorship is merited based on the International Committee of Medical Journal Editors (ICMJE) authorship criteria. A.L.C., W.H.H., A.K., E.J.G., S.K., and H.J.F. contributed to the conception and/or design of the research. A.L.C., W.H.H., J.C., R.G., S.C., and H.J.F. contributed to acquisition of data. M.T.T. and N.Y. contributed to statistical analysis of data. A.L.C., M.T.T., W.H.H., E.J.G., J.S.G., S.K., and H.J.F. contributed to interpretation of data and results. W.H.H. contributed to acquisition of funding. A.L.C., R.G., and H.J.F. contributed to the supervision and management of the research. A.L.C., M.T.T., and H.J.F. contributed to the drafting of the manuscript. All authors contributed to the critical review and revision of the manuscript. M.T.T. and N.Y. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.